WAN Governance

Solving Web applications performance

The ever increasing popularity of web applications is a double-edged sword for many enterprises today. The business side loves that users can be productive from any location with any Internet-enabled device. However, the IT side is often left struggling with network issues and poor performance resulting from the growing complexity and volume of these browser-based apps.

SearchEnterpriseWAN's Jessica Scarpati explores exactly that in "Web application performance: Complex apps make problems for WAN pros." The sheer number of Web and collaboration apps, the fact that many are interconnected or are used as portal to others business applications, their intolerance to packet loss and the competition with the other apps result in timing out, slow performance and finally unusable services. This reduces adoption of Web applications, impacts the productivity and finally affects the image of the IT department.

What strikes here is the incredibly diverse nature of the problems – take just the network, for example: from the intrinsic quality of the transport (please, don't use a network that discards every third packet…), scarce bandwidth in costly locations, competition among users, new cloud-based applications, bandwidth usage explosion, uncertainty about Internet performance, etc.

Applying ad-hoc medicine on such various – and moving – situations will likely lead to disappointing results, costly deployments and fuzzy operations, giving the impression that the IT is pottering about. At Ipanema we believe that this situation can be prevented if we start from the beginning: 1) defining the performance objectives, 2) understanding what's going on, and 3) implementing dynamically the necessary technics to each application connections to guarantee business apps performance and maximize the value of the network to the business.

Believe it or not, this is truly simple: the enterprise just needs to define a few APOs (Application Performance Objectives) for their business applications, and then AQS (Application Quality Score) will automatically show where the performance stands and the autonomic system will automatically control and optimize each user connection to match – or exceed – the defined performance.

Automatic alignment to SLAs across the enterprise WAN alleviates IT — and business — issues of poor performance, usability and overall traffic flows. Building a network and application performance strategy around your users will allow your enterprise to support the complex web applications, user-centric collaboration tools and cloud services that are essential to today's business environment.

Béatrice.

Why it pays to measure application performance

'On a scale of one to ten…' is a classic prompt that has been used to measure just about everything: emotions, the success of a new film or employee performance during reviews. And why not? The scale is concise. It's simple and succinct. It can capture complex data from a variety of sources into condensed, easily transferable and comprehensible numbers.

Now the 'one to ten' scale is helping classify how applications are performing across the WAN via the Application Quality Score (AQS) metric.

AQS provides some basic numbers with big repercussions – numbers that could help businesses build better relationships between IT staff and senior managers, streamline processes, improve customer relations, guide wise investments, minimise downtime, and create some pretty interesting graphs.

What is AQS?

Today there is no standard indicator for application performance across networks – and with good reason: there are very few standard applications. AQS confronts this issue by operating as a composite indicator. By combining a variety of sub-metrics (such as round trip time, server response time, and TCP retransmits) and unique 'one-way' network metrics (such as transit delay, loss and jitter), the 'one to ten' AQS score is a top level view.

How AQS can help improve application performance

AQS acts like a compass, pointing IT departments to where an issue lies and then allowing them to find a solution. Poor application performance can be the result of the network, the server or the application itself. By creating a metric, AQS indicates which of these areas, and where within these areas, any issue is to be found. For example, imagine the AQS indicates that the network is the cause of poor application performance. IT staff would be able to delve deeper. They would be able to answer two related network questions: is the network itself experiencing a fault, or is the network jammed with too much traffic from multiple applications? All this information is contained within the composite parts of the AQS metric.

From there, IT staff could go about finding a solution: if the low AQS score is due to of a faulty network, then the network provider would be responsible for improving it; if the score occurred because of too much traffic over a perfectly suitable network, the IT department would have to unblock the traffic jam. They may choose to do this by determining which applications are business-critical (such as ERP, CRM, videoconference, etc.) and prioritising them. Applications like YouTube or Facebook may be deemed as non-critical, in which case a company may allow an AQS score of 2 or 3. The same company may want an AQS score for critical applications, like SAP or VoIP, to be at 9 or higher.

To keep with the previous example, if the low AQS score indicated the problem was a result of too much business-critical traffic, companies would know they need to direct funds towards increasing bandwidth, and better still, they'd have the information to make the business case.

Why IT people need AQS metrics

Conveying IT information about applications to senior officials can often be challenging. KPI, SLM, SLA, WAN optimisation – these are all terms that managers and senior staff may not be familiar with. AQS helps make the language of IT more approachable and even more visually stimulating, owing to the visual nature of dashboards. The metric of 'one to ten' is simple and easy to understand.

Without such measurements, it's difficult to set application-related goals. Under such circumstances, when issues arise, tension between IT staff and management inevitably follows. Management see IT as expensive and not performing well; the IT department is put in the uncomfortable position of being on the defensive.

In contrast, when there is a metric, a clear indicator that simply and easily points to the heart of an issue, it's much more possible to positively respond to arising difficulties. IT staff have data and information they can provide to senior officials.

Additionally AQS improves user complaint handling. When people do complain, IT staff can see the information needed to help users, even if it is something as simple as being able to describe the problem and tell the customer when it will be fixed. Consider a non-IT example: an individual misses their train. Not knowing when the next train is set to arrive may cause distress. Conversely, knowing how long the delay will be allows the individual to be proactive rather than reactive. Think of the AQS like a digital train board. Using the advanced information on application performance IT departments can better serve users.

(Already posted in Business Cloud News, August 2011)

(Illustration: - Johannes Vermeer, The Geographer 1668-69 oil on canvas; 53×47 cm. Steadelsches Kunstinstitut Frankfurt, Germany)

Box juggling versus integrated solutions

The recent acquisition by Riverbed of Application Delivery Controllers (ADC) Zeus and Aptimize refreshes key questions about the different ways to integrate technologies in enterprises' IT.

On the one hand, one can juxtapose the different required technologies (e.g. WAN optimization, performance monitoring, quality of service, traffic control, web acceleration, etc.). In this vein, Jerry Kennelly (Riverbed's CEO) publicly announced his goal to have the broadest portfolio of performance optimization product. The benefit of such an approach for the vendor is clear: it provides a quick access to a large market that spans above its traditional WAN Optimization green field.

However, it leaves enterprises and their partners (system integrators, MSPs) with the responsibility to manage the complexity, cost and relative inefficiency of juggling with disparate solutions to solve the global issue of application delivery performance.

On the other hand, features integration can be addressed at system standpoint. This is the choice made at Ipanema by tightly coupling key features like Application Visibility, QoS and Control, WAN Optimization and Dynamic WAN Selection in the Autonomic Networking System. This tight integration makes the software design trickier, of course, as many issues have to be addressed (how to protect real-time flows from the other one that you accelerate?) but it proved to be much more efficient as it allows features to cross-fertilize each other (for example application classification can be automatically used for critical applications performance guarantee or intelligent WAN selection).

May be even more importantly, system-level consolidation dramatically simplifies the life of system integrators, managed service providers and enterprises as it prevents them from to taking care of making disparate collections of boxes and software behave coherently, at design level and also during operations.

In a nutshell: integration at box level is simple for vendors and complex for customers. Integration at system level is more complex for vendor while much simpler for customers.